Cloud Native
- Distributed
- Automated
- Organizational
- Anti-fragile
- Replaceable
I like 'cloud native' better than microservice. I think its more descriptive and doesn't have the awkwardness of having 'micro" in the name.
Cloud Native with Spring Boot
It needs to be super easy to implement and update a service.
DEMO
Example Distributed System: Minified
No Man (Microservice) is an Island
It's excellent to be able to implement a microservice really easily
(Spring Boot), but building a system that way surfaces
"non-functional" requirements that you otherwise didn't have.
There are laws of physics that make some problems unsolvable
(consistency, latency), but brittleness and manageability can be
addressed with generic, boiler plate patterns.
Emergent features of cloud native systems
Coordination of distributed systems
leads to boiler plate patterns
- Distributed/versioned configuration
- Service registration and discovery
- Routing
- Service-to-service calls
- Load balancing
- Circuit Breaker
- API Gateway
- Distributed messaging
Example: Coordination Boiler Plate
Netflix OSS
- Eureka
- Hystrix & Turbine
- Ribbon
- Feign
- Zuul
- Archaius
- ...
Spring Cloud
= Spring Boot & Netflix OSS
Example: Spring Cloud and Netflix
Configuration Server
- Pluggable source
- Git implementation
- SVN implementation
- Versioned
- Rollback-able
- Configuration client
auto-configured via starter
DEMO
Declarative Rest Client: Feign
- Pluggable Codecs, Clients and Annotations
- Aims to be simple
- Creates a proxy from a Java interface
DEMO
Hystrix
Hystrix Fallback
Circuit Breaker Metrics
- Via actuator
/metrics
- Server side event stream
/hystrix.stream
- Dashboard app via
@EnableHystrixDashboard
DEMO
How Netflix uses Zuul
- Authentication
- Insights
- Stress Testing
- Canary Testing
- Dynamic Routing
- Service Migration
- Load Shedding
- Security
- Static Response handling
- Active/Active traffic management
DEMO
Spring Cloud Actuator
| Spring Cloud Commons |
Spring Cloud Netflix |
/env (POST) |
/routes (POST) |
/pause (POST) |
/routes (GET) |
/refresh (POST) |
/hystrix.stream/** (GET) |
/restart (POST) |
|
/resume (POST) |
DEMO
Spring Cloud Bus
- Distributed actuator
/bus/env and /bus/refresh actuator endpoints- uses Spring Messaging and Spring Integration
DEMO
No soup for you!
Spring Cloud Security
Enable Single Sign On (SSO) with an OAuth2 provider declared in external properties.
@EnableOAuth2Sso
Enable security using OAuth2 access tokens
@EnableOAuth2Resource
DEMO
Spring Cloud Future
Milestones:
Previews, experiments or ideas (ie: no guarantees!)
Rx Java
- Reactive: push vs. pull
- Functional
- Composable
- Return
Observable from Spring MVC
Controller (soon)
- API Gateway combining services
Spring Restdocs
- Programatically generated snippets from unit tests!
- Hand write asciidoc docs including generated snippets
[source,http]
HTTP/1.1 200 OK
Content-Type: application/hal+json
{
"_links" : {
"users" : {
"href" : "http://localhost:11070/users{?page,size,sort}",
"templated" : true
},
"profile" : {
"href" : "http://localhost:11070/alps"
}
}
}
https://github.com/spring-projects/spring-restdocs
Continuous Delivery
- Microservices lend themselves to continuous delivery.
- You actually need continuous delivery to extract maximum value.
- New: ALM support in Cloudfoundry from Cloudbees
Cloudfoundry
- Environment Provisioning
- On-Demand/Automatic Scaling
- Failover/Resilience
- Routing/Front-end Load Balancing
- Monitoring
Deploying services needs to be simple and reproducible
$ cf push app.groovy
and you don't get much more convenient than that.
(Same argument for other PaaS solutions)
Micro vs Monolithic... is NOT new
From: kt4@prism.gatech.EDU (Ken Thompson)
Subject: Re: LINUX is obsolete
Date: 3 Feb 92 23:07:54 GMT
Organization: Georgia Institute of Technology
I would generally agree that microkernels are probably the wave
of the future. However, it is in my opinion easier to implement
a monolithic kernel. It is also easier for it to turn into a
mess in a hurry as it is modified.
Regards, Ken
What's wrong with a monolith?
